bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1
2 Contribution of transition and stabilization processes to
3 speciation is a function of the ancestral trait state and selective
4 environment in Hakea
5 Byron B. Lamont, Sh-hoob M. El-ahmir, Sim Lin Lim, Philip K. Groom and Tianhua He*
6 Department of Environment and Agriculture, Curtin University, PO Box U1987, Perth, WA
7 6845, Australia;
8 *Author for correspondence: [email protected]
9 ABSTRACT
10 Currently the origin and trajectories of novel traits are emphasised in evolutionary studies, the
11 role of stabilization is neglected, and interpretations are often post hoc rather than as
12 hypothesised responses to stated agents of selection. Here we evaluated the impact of
13 changing environmental conditions on trait evolution and stabilization and their relative
14 contribution to diversification in a prominent Australian genus, Hakea (Proteaceae). We
15 assembled a time-based phylogeny for Hakea, reconstructed its ancestral traits for six
16 attributes and determined their evolutionary trajectories in response to the advent or
17 increasing presence of fire, seasonality, aridity, nectar-feeding birds and (in)vertebrate
18 herbivores/granivores. The ancestral Hakea arose 18 million years ago (Ma) and was broad-
19 leaved, non-spinescent, insect-pollinated, had medium-sized, serotinous fruits and resprouted
20 after fire. Of the 190 diversification events that yielded the 82 extant species analysed, 8−50%
21 involved evolution, stabilization or re-evolution (reversal) of individual novel traits. Needle
22 leaves appeared 14 Ma and increased through the Neogene/Quaternary coinciding with
23 intensifying seasonality and aridity. Spinescence arose 12 Ma consistent with the advent of 1 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
24 vertebrate herbivores. Bird-pollination appeared 14 Ma in response to advent of the
25 Meliphagidae in the early Miocene. Small and large woody fruits evolved from 12 Ma as
26 alternative defenses against granivory. Fire-caused death evolved 14 Ma, accounting for 50%
27 of subsequent events, as fire became less stochastic. Loss of serotiny began in the late
28 Miocene as non-fireprone habitats became available but only contributed 8% of events.
29 Innovation and subsequent stabilization of functional traits promoted the overall species
30 diversification rate in Hakea by 15 times such that only three species now retain the ancestral
31 phenotype. Our approach holds great promise for understanding the processes responsible for
32 speciation of organisms when the ancestral condition can be identified and the likely selective
33 agents are understood.
34 Key words: Ancestral state reconstruction; drought, fire; fruit size, granivore/herbivore,
35 Hakea; insect/bird pollination; leaf shape, Miocene, serotiny, spinescence, resprouting,
36 species diversification, trait proliferation/transition/stabilization
37
38 INTRODUCTION
39 Studies that capture patterns of speciation associated with changes in environmental
40 conditions provide compelling support for the key role of functional trait shifts in the process
41 of evolution by natural selection (Jetz et al., 2012). Natural selection can induce the evolution
42 of novel traits whose fitness exceeds that of the incumbent trait (directional selection) or
43 perpetuation of the current trait whose fitness exceeds that of a former or alternative trait
44 (stabilizing selection) (Lemey et al., 2009). Phylogenetic methods have been developed to
45 investigate a wide range of questions regarding species evolution, including the inference of
46 ancestral traits (He et al., 2011, 2012; Crisp et al., 2011) and to address the relationship
47 between traits and rates of speciation (Litsios et al., 2014). While currently the origin and
2 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
48 evolutionary trajectories of novel traits are emphasised the role of stabilization has been
49 neglected and interpretations have often been post hoc rather than as hypothesised responses
50 to stated agents of natural selection. This is partly because of ignorance of the advent or
51 strength of the postulated selective agents. Significant questions remain: To what extent do
52 directional and stabilizing process contribute to trait proliferation? Do their contributions vary
53 between attributes, traits and/or over geological time? Is the proliferation of a trait at the
54 expense of its alternative traits? Can patterns of directional and stabilizing selection over time
55 be interpreted in terms of the advent or changes in the intensity of particular agents of
56 selection?
57 Theory and concepts
58 Each alternative state of a species attribute is here termed a trait. Increase in occurrence of a
59 given trait through a phylogeny is defined as trait proliferation (He et al., 2011, Lamont et al.,
60 2013). The fraction of total diversification events that result in the presence of that trait is the
61 trait proliferation rate (this may also be given on an absolute basis per unit time, as for species
62 diversification). Trait proliferation results from two evolutionary processes: transition − a new
63 trait arises during the event, and stabilization − the trait is conserved during the event.
64 Transition rate (TR) is the fraction of events in which the trait arises relative to the maximum
65 number in which that trait could occur, while stabilization rate (SR) is the fraction of events
66 in which the trait is retained relative to the maximum number in which that trait could occur.
67 Thus, the (net) trait proliferation rate (PR) = TR + SR. Generally, for a pair of opposing traits,
68 1 and 2, evolving in a clade with a total of n/2 nodes and thus n diversification events for the
69 period of interest, PR1 = TR2/1 + SR1/1, TR2/1 = Σ(2 1 events)/n, and SR1/1 = Σ(1 1 events)/n.
70 Trait reversals are successive transitions that return the phenotype to the previous trait state in
71 the lineage, as an inverse function of the stability through time of the selective pressure for
3 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
72 that trait. The concepts of rates of diversification, proliferation, stabilization and transition,
73 and reversals are illustrated with a concrete example in Fig. 1.
74 A new agent of directional selection usually operates in a different spatial or temporal
75 dimension than the existing agents and becomes a supplementary force that initially retards
76 speciation and then promotes it once an adapted genotype has evolved followed by rampant
77 speciation into the “vacant niche” now available as the new trait proliferates (Lamont et al.,
78 2013). The premise here is that the more habitats (niches) a given area can be divided into,
79 the greater the opportunities for novel genotypes to arise. Accepting that new, alternative
80 traits supplement rather than replace ancestral traits as options, the contribution of novel traits
81 to speciation can be calculated as the inverse of the percentage contribution of the ancestral
82 trait to all subsequent diversification events (Y): speciation promotional rate (SPR) = 100/Y.
83 Thus, the smaller Y, the greater the contribution of novel traits to the subsequent
84 diversification events.
85 Evolution and adaptations of Hakea
86 Nutrient-poor, fire-prone, Mediterranean-type regions with a prolonged hot, dry season and
87 exposed to intensive pressure from pollinators, herbivores and granivores are characterised by
88 high species richness and endemism (Cowling et al., 1996) and should provide suitable
89 scenarios to examine these issues. Hakea is a shrub genus of over 150 species, spread
90 throughout Australia but best represented in mediterranean southwestern Australia, and
91 renowned for its great variation in leaf and fruit morphology, pollinators, climate and fire
92 tolerances and susceptibility to herbivores and granivores (Groom and Lamont, 1996a, 1997,
93 2015; Lamont et al., 2015, 2016; Hanley et al., 2009, Rafferty et al., 2010).
94 Hakea is highly sclerophyllous with needle-leaved and broad-leaved species. Needle leaves
95 are twice as thick as broad leaves implying that they have had greater exposure to drought
4 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
96 and heat (Lamont et al., 2015). Previous molecular analysis indicates that Hakea originated in
97 the early Miocene directly from non-fireprone, rainforest ancestors (Sauquet et al., 2009;
98 Lamont and He, 2012). Hakea most probably originated in the sclerophyll shrublands of
99 southwestern Australia where it continued to diversify strongly until the present (Lamont et
100 al., 2016). From the mid-Miocene, it gradually speciated and migrated onto recently-exposed,
101 rocky substrates, sclerophyll forests and woodlands, and across the drier centre of Australia to
102 the moister margins. Thus, we hypothesise that the ancestral Hakea leaves were broad,
103 reflecting their mesic heritage, that they were retained (or re-evolved) in temperate
104 environments, but that needle leaves arose in the mid-Miocene and proliferated strongly
105 through the late-Miocene to present. Many hakeas have spiny leaves with a sharp apex and/or
106 acute, marginal teeth (Barker et al., 1999). Spiny hakea leaves are more effective at deterring
107 herbivory by kangaroos than broad leaves (Hanley et al., 2007), and are moderately effective
108 at deterring black cockatoos from reaching the woody fruits of hakeas that contain highly
109 nutritious seeds (Groom and Lamont, 2015). Macropods appeared from 17 Ma (Prideaux and
110 Warburton, 2010), soon after the evolution of Hakea. The median stem of black cockatoos
111 (Cacatuidae, Calyptorhynchinae) is positioned at 21.5−15 Ma (White et al., 2011). Needle
112 leaves lend themselves to termination by a sharp apex, so once they appeared selection
113 pressure from vertebrate herbivores/granivores would have promoted the evolution and
114 stabilization of sharp-tipped leaves among vulnerable lineages.
115 Pollinator-driven speciation has been invoked to explain plant richness in some biodiversity
116 hotspots, since pollinator shifts usually provide effective barriers to gene flow, thereby
117 contributing to the origin of new plant lineages (van der Niet et al., 2014). Hanley et al.
118 (2009) concluded that insect pollination was ancestral in Hakea followed by repeated bouts of
119 bird pollination. From their molecular phylogeny of 51 Hakea species, mainly from eastern
120 Australia, Mast et al. (2012) concluded the reverse. Either interpretation is possible since it is
121 now known that honeyeaters (Meliphagidae) originated 23.5 Ma, though they only radiated 5 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
122 strongly from 15 to 5 Ma (Joseph et al., 2014). We hoped to resolve this disagreement by
123 adding more West Australian species to our analysis. We expected bird pollination to
124 stabilize quickly as Toon et al. (2014) showed for bird-pollinated legumes that this was an
125 irreversible process.
126 Hakeas produce woody follicles that vary greatly in size (20−40450 mg) (Groom and Lamont,
127 1997). Their seeds are highly nutritious (Groom and Lamont, 2010). By far the most
128 destructive granivore of hakeas in southwestern Australia is Carnaby's black cockatoo
129 (Calyptorhynchus latirostris, Stock et al. 2013). Two ways of dealing with granivores have
130 been observed: small, camouflaged fruits that are difficult to detect visually and large,
131 exposed fruits that resist attack mechanically (Groom and Lamont, 1997). Thus, we
132 hypothesize that fruit size has taken two directions in Hakea: fruits have become either
133 smaller (and protected or cryptic within spiny foliage, Hanley et al., 2009) or larger and
134 woodier and that these transitions are strongly unidirectional.
135 Recurrent fire is one of the key factors to high species richness in fire-prone ecosystems
136 (Cowling et al., 1996; Simon et al., 2009). Whole-plant responses to fire can be placed into
137 two regeneration syndromes: resprouters that survive fire and recover via dormant buds or
138 meristems protected beneath the bark of trunks or underground organs (Clarke et al., 2013),
139 and nonsprouters that are killed by fire and population regeneration relies solely on seedlings.
140 Both trait-types are well represented among hakeas (Groom and Lamont, 1996b).
141 Nonsprouters live for a shorter time and have a higher fecundity than resprouters (Lamont &
142 Wiens, 2003; Pausas & Verdu, 2005), and are therefore hypothesized to have higher
143 speciation rates (Wells, 1969). Resprouters have an adaptive advantage over nonsprouters
144 when a) fire is either frequent, rare or highly stochastic, b) conditions do not favor growth or
145 seed production e.g. infertile soils or intense competition (fertile soils, high rainfall), and/or c)
146 conditions do not favor seedling recruitment or adult survival (Ojeda, 1998, Lamont et al.,
6 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
147 2011). As the climate became more seasonal through the Miocene, the accumulation of dry
148 matter would have been promoted and fires would have become less stochastic and more
149 likely to occur within the lifespan of the shorter-lived nonsprouters, essential for their
150 promotion (Enright et al., 1998a) but not necessarily at the expense of resprouters (Enright et
151 al., 1998b). Increasing occurrence of arid periods (glacials) would have created mosaics of
152 deep sands and exposed laterites and granites (Glassford and Semeniuk 1995) favoring
153 nonsprouters and resprouters respectively (Lamont and Markey, 1995). Thus, existing
154 knowledge does not enable us to predict the ancestral regeneration strategy of Hakea.
155 Whatever trait was ancestral we can speculate that some of its descendants must soon have
156 transitioned to the other trait, and transition and stabilization rates in both would have been
157 similar throughout the Neogene-Quaternary.
158 Serotiny (on-plant seed storage) is adaptive under conditions that restrict annual seed
159 production (poor soils, low rainfall) in fire-prone environments (for cueing seed release) with
160 a reliable wet season (for effective seedling recruitment) (Lamont et al., 1991, Cowling et al.,
161 2005). While plant death from drought is sufficient to induce seed release among hakeas,
162 postfire conditions are still required for optimal seedling recruitment (Causley et al., 2016).
163 Thus, we postulate that strong serotiny is the ancestral condition in Hakea and that
164 stabilization will be the main evolutionary process for this trait in its subsequent
165 diversification in response to intensifying fire and seasonality. Loss of serotiny will be a later
166 development corresponding to the gradual appearance of fire-free habitats. Hakeas that
167 migrate to summer-rainfall grasslands (savannas) that developed in the late Miocene can also
168 be expected to become nonserotinous.
169 In this study, using a time-based phylogeny for Hakea we assembled (El-ahmir et al., 2015),
170 we reconstructed the ancestral traits for six attributes (with 15 trait states) and determined
171 their evolutionary trajectories in response to the advent or increasing presence of fire,
7 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
172 seasonality, aridity, nectar-feeding birds and vertebrate/herbivores/granivores. We attempted
173 to identify traits of the putative ancestor and the relative contribution of transition and
174 stabilization processes to the frequency of alternative traits over geological time to account
175 for trait representation among the extant species. Our objective was to evaluate the impact of
176 changing environmental conditions on trait evolution and their contribution to diversification
177 in Hakea to provide insights on the factors and processes explaining high species richness in
178 this prominent Australian genus.
179 MATERIALS AND METHODS
180 Phylogenetic reconstruction
181 We built a time-based Hakea phylogeny (El-Ahmir et al., 2015). Briefly, we included 82
182 Hakea species, each with eight gene sequences extracted from NCBI (Mast et al., 2012),
183 combined with new sequences that we generated. The outgroup included Grevillea juncifolia,
184 Finschia chloroxantha and Buckinghamia celsissima and their DNA sequences were obtained
185 from NCBI. We set the calibration point for the origin of the subfamily Grevilleoideae (to
186 which Hakea belongs) at 70.6 Ma based on the fossil Lewalanipollis rectomarginis used by
187 Sauquet et al. (2009). We used BEAST v2.1.0 to estimate phylogeny and divergence times
188 under a strict clock model (Drummond et al., 2006), and further details on the methods are
189 provided in El-Ahmir et al. (2015).
190 Trait data
191 We collated leaf shape and spinescence from Barker et al. (1999), Hanley et al. (2009),
192 personal field observations and database of the State Herbarium of Western Australia
193 (http://www.flora.sa.gov.au). Needle leaves were recognized as rounded in cross-section with
194 a length:width ratio of >20:1. Heteroblastic species, with seedling leaves initially broad
195 becoming needle by the end of the first growing season or seasonally broad to needle (Groom
8 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
196 et al., 1994a), were also identified. Blunt leaves had a mucro or marginal teeth with
197 length:width ratio <1:1 while sharp leaves were >2:1.
198 For pollinator types, Hanley et al. (2009) showed that stigma–nectary distance (SND) in
199 Hakea is a reliable predictor of pollinator class (also adopted by Mast et al. 2012). All known
200 or putative insect-pollinated species have a SND <13 mm and all known or putative bird-
201 pollinated species have a SND >13 mm. This is supported by the shortest bill length of the
202 principal bird pollinators in Australia (family Meliphagidae) is 12 mm (Paton and Ford, 1977)
203 while no known insect pollinator in Western Australia can touch the nectary and pollen
204 presenter simultaneously if the SND >12 mm. We therefore assigned species with SND <13
205 mm to the insect-pollination class and >13 mm to the bird-pollination class. We took SND
206 from Hanley et al. (2009) and Mast et al. (2012). Approximate SND for the remaining
207 species were obtained from pistil lengths in Barker et al. (1999).
208 Fruit size, as dry fruit weight, was obtained from Groom and Lamont (1997). If not available
209 there, the three mean fruit dimensions were obtained from Barker et al. (1999), converted to
210 volume and multiplied by mean fruit density in Groom and Lamont (1997). They were
211 divided into four size classes: <1, 1−5, >5 g, such that the 1−5 g class accounted for about
212 half of species. Postfire response/regeneration strategy was collated from Groom and Lamont
213 (1996b), Barker et al. (1999) and Young (2006). Each species was assigned as either a
214 nonsprouter or resprouter (with two species recognized to have both fire response forms in
215 different populations). Level of serotiny was obtained from Groom and Lamont (1997) and
216 images on the web (especially http://www.flora.sa.gov.au).
217 Trait reconstruction through the phylogeny
218 We used MultiState in BayesTraits (Pagel and Meade 2006) to determine the most likely
219 ancestral traits for the Hakea phylogeny. First, we tested which of the possible models
9 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
220 (simple or complex, associated with uniform rates of 0~30) should be used via the log Bayes
221 factor (log BF) recommended by Pagel and Meade (2006). We excluded morphological data
222 for the outgroup in order to avoid potential biases in trait assignment because they do not
223 adequately represent the associated clades. We applied the best-fit model parameters to our
224 MC tree in a Bayesian framework using MCMC sampling to search for optimal parameter
225 estimates. The MCMC parameter searches consisted of 1,000,000 iterations with 25,000
226 discarded as burn-in. We used maximum likelihood parameter estimates as starting values in
227 the MCMC analyses. We also used the continuous random walk (Model A) associated with
228 the MCMC method to determine whether pairwise traits evolved in a correlated manner, and
229 BayesFactor was used to determine the significance of correlation between any two traits
230 (Pagel and Meade, 2006). Trait reconstruction of fruit size was carried in Mesquite using a
231 parsimonious procedure (Madison et al., 2007).
232 Speciation and trait proliferation rates
233 Net species diversification rate was calculated as (Ni + t − Ni)/(Ni.t), where N is the number of
234 lineages at the start, i, and end, i + t, of the time interval, t (He et al. 2011) for the three
235 geological periods/epochs in which Hakea has been recorded as well as overall. The
236 geological boundaries were set according to the International Commission on Stratigraphy
237 (www.stratigraphy.org), while the start time in the Miocene was set at the time that Hakea
238 first appeared. Following trait assignment to each node of the phylogeny, trait stabilization
239 and transition rates (see Introduction) were determined for the three periods/epochs and
240 overall by counting their number in each time interval. Where the ancestor was ambiguous
241 this event was omitted from the counts as the process was unclear. They were then converted
242 to the fraction that each process contributed to total proliferation within the trait and between
243 all traits of that character. The number of reversals was also noted: i.e. a trait reverting to its
244 immediate preceding trait. Individual speciation promotional rates (SPR) for the three
10 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
245 geological periods were determined from the percentage of events retaining the ancestral trait,
) 246 Y, where SPR = 100/Y (see Introduction). Generally, SPRn =∏$Ͱͥ 100/ where Ys for the n
247 attributes assessed are multiplied to give their total promotional effect on species
248 diversification. SPRn was converted to its fractional contribution to species diversification for
249 the n attributes assessed: (SPRn − 1)/ SPRn.
250 RESULTS
251 Hakea time-based phylogeny
252 The Bayes MCMC analysis indicated that the Hakea stem arose 18.0 Ma [with the 95%
253 highest density probability (HDP) at 15.8−20.2 Ma] and split into two clades (defined as
254 clades A and B in Mast et al. 2012) 14.1 Ma (95% HPD, 12.5–15.8 Ma). The phylogeny was
255 strongly supported by the branch posterior probability where 48 out of 81 branches were ≥
256 0.70. The overall topology of Hakea phylogeny was consistent with that in Mast et al. (2012).
257 Net species diversification rate in the Miocene greatly exceeded that in the Pliocene (9.6×)
258 and Quaternary (13.5×) and the overall rate was dictated by the Miocene rate as it was the
259 longest period (Table S1).
260 Evolutionary trajectories for two leaf attributes
261 Trait reconstruction showed that the most recent ancestor (MRA) had broad leaves (P = 0.61)
262 that were blunt-tipped with smooth margins (P = 0.88) (Fig. 2). The phylogeny split into
263 needle (A) (P = 0.78) and broad (B) (P = 1.00) clades by 14.1 Ma. Heteroblasty arose 6.9 Ma.
264 Both clades remained blunt-tipped (P = 0.69, 1.00) but sharp tips emerged in one branch of
265 the A clade 12.7 Ma. While the transition rate for needle/heteroblastic leaves exceeded that of
266 broad leaves in the Miocene, proliferation of broad leaves accounted for 60% of the
267 diversification events (Table S1A). Broad leaf proliferation continued (mainly through
268 stabilization) at the expense of needle but not of heteroblastic leaves through the Pliocene and 11 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
269 Quaternary. Overall, 65% of total proliferations were of broad leaves (mainly stabilization)
270 with 33 reversals to broad leaves, 30% to needle and 5% to heteroblastic (mainly recent
271 transitions), with the overall transition rate of non-broad leaves 2.6 times broad leaves. The
272 evolution of non-broad leaves increased the overall speciation rate by 54%, greatest in the
273 Miocene. Spiny leaves proliferated at a similar rate as non-spiny leaves in the Miocene but
274 the rate for spiny leaves declined slightly through the Pliocene and Quaternary due to
275 reducing stabilization but increased slightly among non-spiny leaves due to increasing
276 stabilization (Table S1B). Reversals were negligible. The transition rate for spiny leaves was
277 twice that for non-spiny leaves, with their advent and proliferation increasing the
278 diversification rate by 73%.
279 Evolutionary trajectories for two reproductive attributes
280 The MRA of Hakea showed a high posterior probability (P = 0.75) of being insect-pollinated.
281 The basal split of the genus was accompanied by a shift to bird pollination 14.1 Ma in clade B
282 (P = 0.82) but retention of insect pollination in clade A (P = 0.98) (Fig. 3). A reversal
283 occurred in clade B 12.6 Ma while pollination transitioned to birds 12.1 Ma in clade A that
284 remained predominantly insect-pollinated. Overall, 78 reversals occurred (Table S1C). The
285 switch to bird pollination was restricted to the Miocene with transitions accounting for 32%
286 of bird proliferation events, and increasing stabilization through the Pliocene/Quaternary.
287 Bird to insect transitions occurred in the Pliocene but not in the Quaternary. Overall transition
288 rates for insect and bird pollination were similar, with bird pollination accounting for 30% of
289 events and promoting speciation by 41%. The MRA had a high probability (by parsimony) of
290 producing medium-sized fruits (1.0−5.0 g). Smaller (<1.0 g) and larger (>5.0 g) fruits first
291 arose 12.1 Ma in clade A and smaller fruits appeared 6.5 Ma in clade B (Table S1D). In the
292 Miocene, 19% of events involved transitions to other than medium-sized fruits but
293 proliferation of medium-sized fruits predominated. Proliferation of small fruits (46% of
12 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
294 events) dominated in the Pliocene, through both transitions and stabilization, and proliferation
295 of non-medium-sized fruits contributed 150% to the stimulation of diversification events. In
296 the Quaternary and overall, proliferation of medium and non-medium fruits contributed
297 equally to all diversification events. Only medium fruits were sometimes the outcomes of
298 reversals; all other transitions were unidirectional with medium→small accounting for 30
299 events, medium→medium-large/large for 11 events, and medium→medium-large→large for 8
300 events. Overall, 24% of all events involved transitions to non-medium fruits and their
301 proliferation accounted for an 88% increase in the speciation rate.
302 Evolutionary trajectories for two fire-adapted attributes
303 Postfire regeneration of the MRA was via resprouting though the posterior probability was
304 not strong (P = 0.62). The ancestor of clade A was a resprouter (P = 0.73), while clade B was
305 a nonsprouter (P = 0.86) (Fig. 4). By 12.7 Ma, nonsprouters also evolved in clade A. By the
306 end of the Miocene, diversification events were spread almost uniformly between resprouters
307 and nonsprouters (Table S1E). Transitioning to nonsprouters remained strong in the Pliocene
308 but ceased among resprouters. Transitioning ceased in the Quaternary with nonsprouting
309 promoting 140% more speciation through stabilization in that period. Overall, proliferation
310 among resprouters and nonsprouters was similar with the advent of nonsprouters doubling the
311 speciation rate due to similar high rates of stabilization, though transitions to nonsprouting
312 approached twice that for resprouting. Reversals were common among resprouters but only
313 20% of reversals involved nonsprouters. Serotiny was the MRA with P = 1.00 and both major
314 clades remained serotinous (P = 1.00). There was an isolated occurrence of weak/nil serotiny
315 12.1 Ma and five more subsequent origins in clade A but non-serotiny never arose in clade B.
316 Stabilization among moderately/strongly serotinous lineages dominated trait proliferation
317 throughout hakea's history with limited transition to weak/non-serotiny in the Miocene
318 followed by stabilization in the Pliocene and absence of proliferation in the Quaternary.
13 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
319 Overall, stabilization of serotiny was the main process with proliferation of non-serotiny
320 accounting for 7% of events and it increased speciation by 8%. All transitions were
321 unidirectional.
322 Promotion of species diversification
323 The overall speciation promotional rate (SPR6) induced by the advent of novel traits was
324 given by 1.54 × 1.73 × 1.41 × 1.85 × 2.02 × 1.08 = 15.16. Thus, 93.4% of diversification
325 events (ignoring reversals) can be attributed to the presence of at least one non-ancestral trait.
326 Three species possessed the six ancestral traits (H. candolleana, H. ceratophylla, H.
327 eriantha), all in the same subclade of 14 species, but they included two reversals
328 (nonsprouter→resprouter, medium-large→medium fruits). Thus, 96.3% of extant species lack
329 at least one ancestral trait. One species (H. divaricata) had five of six traits in the advanced
330 condition.
331 Correlated evolution between traits
332 Correlation analysis using the BayesFactor (BF) showed no relationship between any pairs of
333 attributes (BF < 1.0) except leaf shape and spinescence, with needle leaves more likely to be
334 spiny (BF = 5.3).
335 DISCUSSION
336 Trait reconstruction of the ancestral Hakea phenotype shows it to have been broad-leaved,
337 non-spinescent and insect-pollinated, with medium-sized, serotinous fruits and resprouting
338 after fire. Resprouting and serotiny confirm that the associated vegetation was fireprone and
339 experienced a reliable postfire wet season by 18 Ma (Lamont et al., 2013). It is clear that
340 Hakea changed radically at the level of fire-related adaptations, including woodiness of their
341 fruits, when migrating from the non-fireprone environment of its ancestors (nonsprouting and
14 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
342 nonserotinous), whereas leaf form (broad, non-spinescent) and reproductive biology [insect-
343 pollinated, medium-sized (1−5 g) fruits] were initially conserved. Nevertheless, within 4 My,
344 two (sub)clades had evolved with quite different syndromes of traits: one (A) that retained
345 resprouting but possessed needle leaves many of which developed sharp apices, was bird-
346 pollinated and where the largest woody fruits (>10 g) were produced, and the other (B) that
347 became nonsprouting but all other attributes were dominated by their ancestral traits. The
348 final outcome was almost equal representation of broad and needle leaves and spiny and blunt
349 leaves, significant presence of bird-pollination, almost equal representation of small (<1 g)
350 and large (>5 g) fruits, equal representation of resprouting and nonsprouting, and limited
351 presence (10%) of weak/non-serotiny. Only three of 82 extant species retain all six ancestral
352 traits and even two of these traits were the outcome of reversals. At the genus level, of 15
353 possible pairs of correlated evolution between attributes, only needle and sharp-pointed
354 leaves were associated through time (attributable to their morphological links).
355 The species diversification rate of Hakea was highest by far in the Miocene than in the more
356 recent epochs. The Miocene was a period of great climatic upheavals and the speciation rates
357 among banksias in Australia (He et al., 2011) and proteas in South Africa (Lamont et al.,
358 2013) (both genera also in Proteaceae) were also an order of magnitude higher then. The same
359 pattern applies to proliferation of traits, with all alternative traits of the six examined highest
360 in the Miocene (obtained by multiplying the percentage contribution to species diversification
361 of each trait by the diversification rate on a time basis).
362 Transition versus stabilization processes
363 While trait initiation (transition) is a vital step in speciation its incorporation into the clade
364 (stabilization) is just as important. That proliferation of a trait through the phylogeny is rarely
365 a function of the transition rate is strongly supported here. Taking leaf shape as an example,
366 the transition from broad to needle leaves overall occurred at >2.6 times the rate as the 15 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
367 reverse transition, yet stabilization of broad leaves occurred at 2.3 times the rate as needle
368 leaves. The net result was the proliferation of broad leaves at 1.85 times the rate of needle
369 leaves because the ratio of stabilization to transition events among broad leaves was five
370 times the rate for needle leaves. In theory, only one initiation step is required for
371 incorporation of a new trait into the clade provided it stabilizes quickly and is not subject to
372 reversals. Thus, the ratio of stabilisation to transition events is a function of the strength of
373 directional selection. The invasion of the savanna grasslands by Protea is a rare example of
374 unidirectional selection associated with a single transition followed by almost universal
375 stabilization (Lamont et al., 2013). In practice, the same trait arises numerous times through
376 the phylogeny while reversals depend on the trait. For Hakea leaf shape, 77% of the 43-
377 recorded reversals were for the recovery of broad from needle leaves rendering transitions to
378 needle less effective and reflecting unstable selective forces. The relative contribution of
379 transition and stabilization events to proliferation depends on both the trait and the time
380 period under consideration.
381 Evolutionary trajectories for leaves
382 By the time Hakea separated from its non-fireprone ancestors 18 Ma, Australia (as much of
383 the world) was experiencing declining levels of rainfall, temperatures and metabolically
384 active atmospheric gases, and increasing seasonality. In addition, the opening up of the
385 vegetation would have exposed them to high light intensity and diurnal temperatures (Jordan
386 et al., 2005) compared with closed forests. If needle leaves increase fitness to such
387 constraints, and currently they account for 45% of species so this genus has a strong
388 propensity to produce them, they should have evolved early in its history and proliferated
389 through stabilization. Indeed, within 4 My, a needle-leaved clade (A) had arisen with strong
390 stabilization leading to 56% of its extant species being needle-leaved. Evolution of needle
391 leaves was greatly delayed in clade B and was mainly expressed through the appearance of
16 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
392 heteroblastic species over the last 5 My (all from broad-leaved ancestors). The latter appeared
393 so recently that there have been no opportunities for reversals unlike needle leaves where
394 reversals to broad have been frequent. These reversals confirm the lability of leaf form among
395 isobilateral leaves as demonstrated ontogenetically by H. trifurcata whose juvenile leaves are
396 needle, a few becoming broad at the start of the growing season in adult plants and needle
397 again as the dry summer approaches (Groom et al., 1994a).
398 The dominance of broad leaves and reversals to them require some explanation. Clearly, leaf
399 form is not the only way of dealing with drought, such as deep root systems (Groom and
400 Lamont, 2015), while broad leaves among hakeas are still highly sclerophyllous (Lamont et
401 al., 2015) with thick cuticles, sunken stomata and a tannin-filled hypodermis (Jordan et al.,
402 2005). They are often narrow or strap-shaped rather than truly broad, such as H.
403 grammatophylla in the 'deadheart' of Australia, and all are vertically oriented. In addition,
404 broader-leaved species have retreated to the moister parts of the landscape or subregions
405 (Groom and Lamont, 1996a). In fact, the frequent reversals in both directions are consistent
406 with climatic oscillations that became characteristic of the Pliocene/Quaternary and their
407 evolutionary tracking.
408 While broad leaves may be spinescent, such as H. cristata, needle leaves that are already rigid
409 and with a sclerified apex can readily be transformed into strongly piercing structures through
410 elongation and thinning of the mucro. This morphogenetic link explains the unique
411 evolutionary correlation through time of needle and spinescent leaves. Thus, following a
412 small delay, one branch of the A clade became spinescent at 12.7 My. The broad-leaved B
413 clade remained essentially non-spinescent. If spinescence is effective against herbivores
414 (macropods) and florivores/granivores (emus, cockatoos) (Hanley et al., 2007) the delay in its
415 appearance cannot be attributed to their absence as all were present by this time but they
416 speciated gradually and their selective effects would have intensified over time. It is of
17 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
417 interest that transitioning to spiny leaves was most marked in the Quaternary, a time when
418 modern cockatoos evolved in SW Australia, though their ancestors were present from the
419 early Miocene (Joseph et al., 2014). Reversals were negligible indicating strong directional
420 selection. Why more events did not yield spiny leaves is partly attributable to morphological
421 constraints, the fact that all Hakea leaves are highly unpalatable and not grass-like (Rafferty
422 et al., 2010), and ability of vertebrates to learn to overcome physical deterrents (Hanley et al.,
423 2007).
424 Evolutionary trajectories for flowers and fruits
425 Bird-pollinated flowers evolved from insect-pollinated flowers with the split of the genus
426 14.1 Ma. This resolves the disagreements over which was the ancestral condition (Hanley et
427 al., 2009; Mast et al., 2012) caused by misidentifying the basal lineages or not including
428 sufficient (representative) insect-pollinated lineages from SW Australia where the clade most
429 probably arose (Lamont et al., 2016). Honeyeaters (Meliphagidae) were already present in
430 Australia at the time Hakea originated, but these birds only diversified strongly in the mid to
431 late-Miocene, especially among such major pollinators as Phylidonyris, Anthochaera,
432 Lichmera and Lichenostomus (Joseph et al., 2014). In fact, apart from H. cucullata in the
433 Quaternary, the only time flowers increased their size to accommodate bird pollinators was in
434 the Miocene. For the Pliocene/Quaternary it was stabilization processes only. The greater
435 levels of stabilization among insect-pollinated lineages throughout their history explains their
436 current greater abundance and suggests that they have been a greater selective force,
437 possibility associated with their greater reliability rather than morphological diversity that
438 would have favored greater transition rates.
439 Of note are the transitions from bird to insect pollination in the Miocene/Pliocene and the
440 large number of reversals (78), 60% of which were insect bird insect. This is significant on
441 two counts: trait reversibility and fluctuating selection. Bird-pollination is regarded as an 18 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
442 evolutionary 'dead-end' because specialization of floral structures for birds may be
443 irreversible (Toon et al., 2014) so shifts from bird to insect pollination are rare (van der Niet
444 & Johnson 2012). This hypothesis derives from hummingbird pollination systems that require
445 specialised floral structures (Tripp and Manos, 2008). However, specialised floral structures
446 are not essential for honeyeaters because they are generalist pollinators and not obligate
447 nectarivores. This lack of specialization in the honeyeater pollination system implies minimal
448 floral structure specialization – simple elongation of the pistil is sufficient (Hanley et al.,
449 2009). As a result, reversal shifts from bird to insect are possible in situations when bird
450 pollinators become scarce. The evolutionary tracking of the great climatic fluctuations, with
451 their profound effects on the abundance of both birds and insects, that occurred from the mid-
452 Miocene can explain the remarkable number of reversals in both directions. Such great
453 flexibility in pollinator shifts may provide insights in explaining the mechanisms that
454 promoted explosive speciation in Hakea but even more so in its younger sister, Grevillea with
455 150 of 362 species being bird-pollinated (Ford et al., 1979).
456 Morphological variation in fruit structure among hakeas is exceptional and transition rates
457 away from the ancestral medium-sized ancestor equalled their stabilization rates through each
458 of the three epochs. Two groups of granivores, insects and parrots, were already present when
459 Hakea emerged. The transition to small (camouflaged) fruits, which appear better protected
460 against insects than cockatoos (B Lamont, unpublished), was strongest in the Miocene and
461 continued throughout the period, by far the highest among all traits we assessed. Large woody
462 fruits, particularly effective against cockatoos (Groom and Lamont, 2015), followed a similar
463 path. There were no reversals among small and large fruits indicating strong unidirectional
464 selection for these extremes.
465 Evolutionary trajectories for fire-related traits
466 The twin ancestral traits of resprouting (adaptive in the presence of severe, periodic 19 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
467 disturbances where recruitment opportunities are limited) and serotiny (ensuring seeds are
468 released when those limited recruitment opportunities are optimal) demonstrate that Hakea
469 arose in a fireprone environment. We can also surmise that fires were of overall moderate
470 frequency (>5−45 y intervals) but highly stochastic. If fires were at the high frequencies
471 associated with savanna grasslands (<5 y intervals) the plants would have stayed
472 nonserotinous (Lamont et al. 2013). If fire intervals exceeded their longevity then, upon death
473 in the absence of fire, serotinous seeds would have been released onto a hostile seedbed and
474 rarely yielded recruits (Causley et al., 2016). However, within 4 My, a nonsprouting clade (B)
475 had arisen. The outcome was strong transitioning to both fire-response types in the Miocene
476 and all epochs were dominated by stabilizations against a background of almost universal
477 proliferation by serotinous descendents. Of note are both the continuing transitions to
478 nonsprouting in the Pliocene and its steady increase in stabilization rate throughout Hakea's
479 history. It is likely that the trend of increasing aridity and seasonality and declining
480 atmospheric oxygen and carbon dioxide (He et al., 2012) led to less frequent, but more
481 reliable, fires and promotion of nonsprouting (Lamont et al., 2011, 2013). Today, resprouters
482 are better represented in the more fireprone, strongly seasonal northern sandplains of SW
483 Australia and nonsprouters in the less fireprone parts, especially on deeper soils where
484 recruitment and adult survival are more likely (Lamont and Markey, 1995; Groom and
485 Lamont, 1996a).
486 Transitions to non/weak serotiny were rare but five of the six independent origins were in the
487 late Miocene and proliferation was restricted to the Pliocene. Explanations vary but include
488 migrations to frequently-burnt savanna grasslands (H. divaricata lineage) or rarely-burnt
489 aridlands (H. recurva), exposure of novel firefree rock outcrops to which some species
490 adapted (H. clavata), and presence in forests with reliable winter rains ensuring recruitment
491 interfire (H. trifurcata lineage) (Hanley and Lamont, 2001). This pattern has limited parallels
492 with Protea in South Africa where one transition to nonserotiny in grasslands was followed 20 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
493 by increasingly extensive stabilization there that failed to occur in Australian grasslands.
494 However, in both super-regions stabilization of the ancestral condition, serotiny, was by far
495 the dominant process and reversals were negligible. By contrast, the well known lability in
496 whole-plant fire responses (He et al., 2011) was expressed in both Protea and Hakea,
497 although 80% of reversals were to resprouting in Hakea in the Miocene, perhaps reflecting
498 periods of increased or more stochastic fire frequencies as the clade moved to other parts of
499 Australia. The historic levels of fire frequency coupled with the severity of seasonal droughts
500 serve well to interpret the relative abundances and distributions of resprouters and
501 nonsprouters among hakeas, ericas and proteas (Ojeda, 1998; Lamont et al., 2013).
502 Promotion of speciation by non-ancestral traits
503 There are two ways of considering the role of traits in speciation. One is to compare how they
504 have contributed to all diversification events and the other is to estimate to what extent non-
505 ancestral traits have promoted these diversification events. The latter is hypothesis-driven and
506 based on the premise that without trait innovation and subsequent stabilization these
507 diversification events would not have occurred. Thus, it relies on being able to identify both
508 the ancestral state and its pathway through the phylogeny in order to ascertain trait reversals.
509 This means that any extinction events cannot be incorporated into the analysis, but we argued
510 earlier (see Introduction) that most new traits are adjunct to (as they are spatially displaced),
511 rather than replace, ancestral traits anyway. Nevertheless, entire but unknown lineages
512 characterized by certain historically maladapted traits may be missing (or only represented by
513 long branches in the chronogram) if there have been radical environmental shifts that no
514 amount of adjustment, in the absence of fossil evidence, can correct for. However, there is
515 little evidence of extinctions as a significant evolutionary process in the SW Australian flora
516 over the last 10 My (Hopper, 2009).
21 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
517 For the six Hakea attributes examined, proliferation of the non-ancestral trait promoted
518 speciation by 1.08 (weakly serotinous) to 2.02 (fire-killed) times. Overall, this increased
519 speciation by 15.2 times, equivalent to 93.4% of diversification events. Two of the ancestral
520 traits possessed by the three species of identical phenotype to the putative original phenotype
521 were the result of reversals so that even these were the outcome of trait diversification when
522 the ancestral condition would have been temporarily lost. Reversals were particularly
523 prevalent among pollination types and fire-response types but absent altogether from the
524 serotiny types. They represent the net effect of a) the constancy of directional selection, b) the
525 lability of opposing traits, and c) time available for further transitions to occur. Certainly, the
526 high level of lability among fire-response types is consistent with previous studies on Protea
527 (Lamont et al. 2013), Banksia (He et al., 2011) and Restionaceae (Litsios et al., 2014).
528 Clearly, Hakea’s remarkable genetic/morphological malleability in the face of these strong
529 selective agents has resulted in an exceptionally diverse clade and its distribution throughout
530 Australia. We might now wonder to what extent the cumulative contributions by transitions
531 and stabilizations to trait proliferation among critical plant attributes, in response to an array
532 of environmental constraints introduced in the Miocene, have led to the explosive radiation of
533 such speciose genera as Grevillea, Acacia, Melaleuca and Eucalyptus, that currently
534 dominate Australia's sclerophyll flora, and the floras of many other parts of the world subject
535 to similar selective forces.
536 Acknowledgments
537 This study was supported by the Australian Research Council (DP120103389 and
538 DP130103029).
539 References
540 Barker RM, Haegi L, Barker WR. 1999. Hakea. In: Wilson A. ed. Flora of Australia, Volume 17B,
22 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
541 Proteaceae 3, Hakea to Dryandra. Canberra: ABRS/CSIRO. 31-170.
542 Causley CL, Fowler WM, Lamont BB, He T. 2016. Fitness of serotiny in fire- and drought-
543 prone environments. Plant Ecology (in press)
544 Clarke PJ, Lawes MJ, Midgley JJ, Lamont BB, Ojeda F, Burrows GE, Enright NJ, Knox KJE.
545 2012. Resprouting as a key functional trait: how buds, protection and reserves drive
546 persistence after fire. New Phytologist 197: 19-35.
547 Cowling RM, Rundel PW, Lamont BB, Arroyo MK, Arianoutsou M. 1996. Plant diversity in
548 Mediterranean-climate regions. Trends in Ecology and Evolution 11: 362-366.
549 Cowling RM, Ojeda J, Lamont BB, Rundel PW. Lechmere-Oertel R. 2005. Rainfall reliability, a
550 neglected factor in explaining convergence and divergence of plant traits in fire-prone
551 mediterranean-climate ecosystems. Global Ecology and Biogeography 14: 509-519.
552 Crisp MD, Burrows GE, Cook LG, Thornhill AH. Bowman DMJS. 2011. Flammable biomes
553 dominated by eucalypts originated at the Cretaceous-Palaeogene boundary. Nature
554 Communications 2: 193.
555 Drummond AJ, Ho SYW, Phillips MJ, Rambaut A. 2006. Relaxed phylogenetics and dating with
556 confidence. PloS Biology 4: 699-710.
557 El-ahmir SM, Lim SL, Lamont BB, He T. 2015. Seed size, fecundity and postfire
558 regeneration strategy are interdependent in Hakea. PLoS ONE 10: e0129027.
559 Enright NJ, Marsula R, Lamont BB, Wissel C. 1998a. The ecological significance of canopy
560 seed storage in fire-prone environments: a model for nonsprouting shrubs. Journal of
561 Ecology 86: 946-959.
562 Enright NJ, Marsula R, Lamont BB, Wissel C. 1998b. The ecological significance of canopy
23 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
563 seed storage in fire-prone environments: a model for resprouting shrubs. Journal of
564 Ecology 86: 960-973.
565 Ford HA, Paton DC, Forde N. 1979. Birds as pollinators of Australian plants. New Zealand Journal
566 of Botany 17: 509-519.
567 Glassford DK, Semeniuk V. 1995. Desert-aeolian origin of late Cenozoic regolith in arid and semi-
568 arid southwestern Australia. Palaeogeography, Palaeoclimatology, Palaeoecology 114:131–
569 166.
570 Groom PK. Lamont BB. 2015. Plant Life of Southwestern Australia – Adaptations for Survival.
571 Warsaw, Poland: De Gruyter Open.
572 Groom PK. Lamont BB. 1996a. Ecogeographical analysis of Hakea (Proteaceae) in south-western
573 Australia, with special reference to leaf morphology and life form. Australian Journal of
574 Botany 44: 527-542.
575 Groom PK, Lamont BB. 1996b. Reproductive ecology of non-sprouting and re-sprouting Hakea
576 species (Proteaceae) in southwestern Australia. In Hopper SD, Harvey M, Chappill J, George
577 AS. eds. Gondwanan Heritage: Evolution and Conservation of the Western Australian Biota.
578 (Chipping Norton): Surrey Beatty, 239-248.
579 Groom PK. Lamont BB. 1997. Fruit-seed relations in Hakea: serotinous species invest more dry
580 matter in predispersal seed protection. Australian Journal of Ecology 22: 352-355.
581 Groom PK Lamont BB. 2010. Phosphorus accumulation in Proteaceae seeds: a synthesis. Plant and
582 Soil 334: 61-72.
583 Groom PK, Lamont BB. Kupsky L. 1994a. Contrasting morphology and ecophysiology of co-
584 occurring broad and terete leaves in Hakea trifurcata (Proteaceae). Australian Journal of
585 Botany 42: 307-320. 24 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
586 Hanley ME, Lamont BB, Armbruster WS. 2009. Pollination and plant defense traits co-vary in
587 Western Australian Hakeas. New Phytologist 182: 251-260.
588 Hanley ME, Lamont BB, Fairbanks MM, Rafferty CM. 2007. Plant structural traits and their role in
589 anti-herbivore defence. Perspective in Plant Ecology and Evolution 8: 157-178.
590 Hanley ME, Lamont BB. 2001. Herbivory, serotiny and seedling defence in Western Australian
591 Proteaceae. Oecologia 126: 409-417.
592 He T, Lamont BB, Downes KS. 2011. Banksia born to burn. New Phytologist 191: 184-196.
593 He T, Pausas JG, Belcher CM, Schwilk DW, Lamont BB. 2012. Fire-adapted traits of Pinus arose in
594 the fiery Cretaceous. New Phytologist 194: 751-759.
595 Hopper SD. 2009. OCBIL theory: towards an integrated understanding of the evolution,
596 ecology and conservation of biodiversity on old, climatically buffered, infertile
597 landscapes. Plant and Soil 322: 49–86.
598 Jetz W, Thomas GH, Joy JB, Hartmann K. Mooers AO. 2012. The global diversity of birds in space
599 and time. Nature 491: 444-448.
600 Jordan GJ, Dillon RA, Weston PH. 2005. Solar radiation as a factor in the evolution of
601 scleromorphic leaf anatomy in Proteaceae. American Journal of Botany 92: 789–796.
602 Joseph L, Toon A, Nyari AS, Longmore NW, Rowe KMC, Haryoko T, Trueman J, Gardne JL. 2014.
603 A new synthesis of the molecular systematics and biogeography of honeyeaters
604 (Passeriformes: Meliphagidae) highlights biogeographical and ecological complexity of a
605 spectacular avian radiation. Zoologica Scripta 43: 235-248.
606 Lamont BB, Groom PK. 1998. Seed and seedling biology of the woody-fruited Proteaceae.
607 Australian Journal of Botany 46: 387-406.
25 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
608 Lamont BB, He T. 2012. Fire-adapted Gondwanan Angiosperm floras arose in the Cretaceous. BMC
609 Evolutionary Biology 12: 223.
610 Lamont BB, Markey A. 1995. Biogeography of fire-killed and resprouting Banksia species in
611 southwestern Australia. Australian Journal of Botany 43: 283-303.
612 Lamont BB, Wiens D. 2003. Are seed set and speciation rates always low among species that
613 resprout after fire, and why? Evolutionary Ecology 17: 277-292.
614 Lamont BB, Enright NJ, He T. 2011. Fitness and evolution of resprouters in relation to fire. Plant
615 Ecology 212, 1945-1957.
616 Lamont BB, Groom PK, Williams M, He T. 2015. LMA, density and thickness: recognizing different
617 leaf shapes and correcting for their nonlaminarity. New Phytologist 207: 942–947.
618 Lamont BB, He T, Downes KS. 2013. Adaptive responses to directional trait selection in the
619 Miocene enabled Cape proteas to colonize the savanna grasslands. Evolutionary Ecology 27:
620 1099-1115.
621 Lamont BB, He T, Lim SL. 2016. Hakea, the world's most sclerophyllous genus, arose in
622 southwest Australian heathland and diversified throughout Australia over 10 million
623 years. Australian Journal of Botany (in press).
624 Lamont BB, Le Maitre D, Cowling RM, Enright NJ. 1991. Canopy seed storage in woody
625 plants. The Botanical Review 57: 277-317.
626 Lemey P, Salemi M, Vandamme AM. 2009. The Phylogenetic Handbook. Cambridge University
627 Press. Cambridge.
628 Litsios G, Wuest RO, Kostikova A, Forest F, Lexer C, Linder HP, Pearman PB, Zimmermann NE,
629 Salamin N. 2014. Effects of a fire response trait on diversification in replicated radiations.
26 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
630 Evolution 68: 453-465.
631 Maddison WP, Midford PE, Otto SP. 2007. Estimating a binary character's effect on speciation and
632 extinction. Systematic Biology 56:701–710.
633 Mast AR, Milton EF, Jones EH, Barker RM, Barker WR, Weston PH. 2012. Time-calibrated
634 phylogeny of the woody Australian genus Hakea (Proteaceae) supports multiple origins of
635 insect-pollination among bird-pollinated ancestors. American Journal of Botany 99: 472-487.
636 Ojeda F. 1998. Biogeography of seeder and resprouter Erica species in the Cape Floristic
637 Region—Where are the resprouters? Biological Journal of the Linnean Society 63:
638 331–347.
639 Pagel M. Meade A. 2006. Bayesian analysis of correlated evolution of discrete attributes by
640 reversible-jump Markov chain Monte Carlo. American Naturalist 167: 808-825.
641 Paton DC. Ford HA. 1977. Pollination by birds of native plants in southern Australia. Emu 77: 73-85.
642 Pausas JG. Verdu M. 2005. Plant persistence traits in fire-prone ecosystems of the Mediterranean
643 basin: a phylogenetic approach. Oikos 109: 196-202.
644 Prideaux GJ, Warburton NM. 2010. An osteology-based appraisal of the phylogeny and
645 evolution of kangaroos and wallabies (Macropodidae: Marsupialia). Zoological
646 Journal of the Linnean Society 159, 954-987.
647 Rafferty CM, Lamont BB, Hanley ME. 2010. Herbivore feeding preferences in captive and wild
648 populations. Austral Ecology 35, 257–263.
649 Sauquet H, Weston PH, Anderson CL, Barker NP, Cantrill DJ, Mast AR. et al. 2009. Contrasted
650 patterns of hyperdiversification in Mediterranean hotspots. Proceedings of National Academy
651 of Sciences USA 106: 221-225.
27 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
652 Simon MF, Grether R, de Queiroz LP, Skema C, Pennington RT, Hughes CE. 2009. Recent assembly
653 of the Cerrado, a neotropical plant diversity hotspot, by in situ evolution of adaptations to
654 fire. Proceedings of National Academy of Sciences USA 106: 20359-20364.
655 Stock WD, Finn H, Parker J, Dods K. 2013. Pine as fast food: foraging ecology of an
656 endangered cockatoo in a forestry landscape. PLoS ONE 8:e61145.
657 Toon A, Cook LG, Crisp MD. 2014. Evolutionary consequences of shifts to bird-pollination in the
658 Australian pea-flowered legumes (Mirbelieae and Bossiaeeae). BMC Evolutionary Biology
659 14: 43.
660 Tripp EA. Manos PS. 2008. Is floral specialization an evolutionary dead-end? Pollination system
661 transitions in Ruellia (Acanthaceae). Evolution 62: 1712-1736.
662 Van der Niet T, Peakall R, Johnson SD. 2014. Pollinator-driven ecological speciation in plants: new
663 evidence and future perspectives. Annals of Botany 113: 199-211.
664 Wells PV. 1969. The relation between mode of reproduction and extent of speciation in woody
665 genera of the California chaparral. Evolution 23: 264-267.
666 White NE, Matthew J, Phillips MJ, Gilbert MTP, Alfaro-Núñez A, Eske Willerslev E, Mawson PR,
667 Spencer PBS, Bunce M. 2011. The evolutionary history of cockatoos (Aves: Psittaciformes:
668 Cacatuidae). Molecular Phylogenetics and Evolution 59:615–622.
669 Young JA. 2006. Hakeas of Western Australia: A field and identification guide. JA. Young, Perth,
670 Australia.
671
28 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
672 Figures
673
!!!! !! T # 1 A# # T1###################3# stabilization# 1# stabilization########4# T # 1 B# T1# # stabilization# # # T # 1 C# 2# T2####################5# #transition# # #########transition#######6# T # 2 D# # # stabilization# a# b# c# # 0# Time# X ! ! Species#diversification#rate:#(Spp#@#c#−#Spp#@#a)/[Spp#@#a#x#(a#+#b#+#c)]#=#(4#−#1)/(1#x#X)# # Six#diversification#events#over#time#X: # Trait# Genetic#process## Within# Between# Trait#proliferation# (state)# involved# trait# traits# rate# 1# 1 1,#stabilization#rate# 3/4# 3/6# 4/6# 1# n 1,#transition#rate# 1/4# 1/6# 2# 2 2,#stabilization#rate# 1/2# 1/6# 2/6# 2# n 2,#transition#rate## 1/2# 1/6# # where#n#=#one#or#more#alternative#states#to#the#evolved#one.# Speciation#promotory#rate#(increase#in#species#diversification#rate#due#to#the#presence# of#novel#trait#2)#is#given#by#1/(4/6)#=#1.50.#
674
675 Fig. 1. Hypothetical phylogeny showing the evolution of four species from six diversification
676 events with proliferation of two alternative traits, 1 (ancestral) and 2 (novel), of a given
677 character over time, due to both stabilization and transition processes. Diversification rate is
678 relative to the starting number of species/lineages and the time interval, while proliferation,
679 stabilization and transition rates are relative to the maximum number of events in which that
680 trait could occur. Note that the two transitions that yield Sp C is an example of a reversal (to
681 T1).
682
683
29 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
684
685
686 Figure 2. Reconstruction of leaf morphology traits through time in the genus Hakea. Left: leaf
687 shape, broad, needle or heteroblastic (broad followed by needle). Right: leaf spinescence,
688 blunt/nil or sharp apices or teeth.
689
30 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
690
691
692 Figure 3. Reconstruction of reproductive biology traits through time in the genus Hakea. Left:
693 insect or bird pollinated. Right: four classes of fruit size by weight.
694
31 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
695
H. strumosa H. psilorrhyncha H. verrucosa Leaf shape H. florida Spinescence H. varia H. lissocarpha H. cristata needle H. megalosperma sharp apex/margins H. horrida H. ilicifolia broad H. drupacea H. nitida blunt apex/margins H. linearis H. ruscifolia heteroblastic H. prostrata H. auriculata H. spathulata H. denticulata H. preissii H. clavata H. salicifolia H. circumalata H. recurva H. constablei H. propinqua H. sericea H. longiflora H. commutata H. purpurea H. pandanicarpa H. crassifolia H. obliqua H. incrassata H. candolleana H. lasianthoides H. trifurcata H. megadenia H. teretifolia H. ceratophylla H. brownii H. baxteri H. flabellifolia H. eriantha H. arborescens H. persiehana H. brachyptera H. orthorrhyncha H. platysperma H. divaricata H. lorea H. subsulcata H. invaginata H. victoria H. corymbosa H. sulcata H. gilbertii H. grammatophylla H. laurina H. lehmanniana H. obtusa H. marginata H. eneabba H. petiolaris H. cygna needlei H. cygna cygna H. scoparia H. ulicina H. meisneriana H. dactyloides H. conchifolia H. smilacifolia H. stenophylla H. pycnoneura H. cucullata H. anadenia H. hastata H. ferruqinea H. multilineata H. bucculenta H. francisiana H. archaeoides H. trineura Grevillea juncifolia Finschia chloroxantha 20 15 10 5 0 Ma 0 5 10 15 20
696
697 Figure 4. Reconstruction of fire-adapted traits through time in the genus Hakea. Left:
698 nonsprouter (seedlings only) or resprouter. Right: strongly or weakly/nil serotinous.
699
32 bioRxiv preprint doi: https://doi.org/10.1101/207373; this version posted October 23, 2017. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
700
Supplementary materials
Table S1. Paired trait evolution in Hakea apportioned among stabilization (trait retained during diversification event) and transition (trait attained during diversification event) processes in each Epoch based on the molecular chronogram reported here. All node-to-node steps in the phylogeny were treated as diversification events. Where the ancestor was ambiguous this event was omitted from the counts as the process was unclear. Reversals refer to transitions back to the previous trait.
Epoch/period: Miocene (18—5.3 Ma) Pliocene (5.3—2.6 Ma) Quaternary (2.6—0 Ma) Overall (18—0 Ma)
Species diversification rate: 3.23 0.23 0.07 4.48
Trait that Genetic Within Between Within Between Within Between Within Between Number of evolved process trait (%) traits (%) trait (%) traits (%) trait (%) traits (%) trait (%) traits (%) reversals
A. Leaf shape
Broad leaves Stabilization 93.9 56.8 96.9 64.6 94.1 72.7 94.9 61.6 (ancestral) Transition 6.1 3.7 3.1 2.1 5.9 4.6 5.1 3.3 33 Needle leaves Stabilization 87.1 33.3 83.3 20.8 50.0 4.6 84.4 25.2 Transition 12.9 4.9 16.7 4.2 50.0 4.6 15.6 4.6 10 Heteroblasty Stabilization 0.0 0.0 50.0 4.2 0.0 0.0 25.0 1.3 Transition 100.0 1.2 50.0 4.2 100.0 13.6 75.0 4.0 0 100.0 100.0 100.0 100.0
Fisher Test (P, 2-tailed) 0.0367 0.0120 0.0013 <0.0001 Speciation promotion rate 1.65 1.50 1.29 1.54
701
702
33